Analysis of Wind Turbine Operation Behavior Based on Clustering Algorithm

Abstract

This research offers an innovative approach to the analysis of wind turbine operation behaviors through the use of clustering algorithms. Fundamental steps of data preprocessing were undertaken, including comprehensive data cleaning and feature selection from an open-source wind energy dataset. A comparative study of numerous clustering algorithms was then conducted, with the findings indicating that hierarchical clustering provides an optimal method for extracting wind turbine behavior patterns across diverse time dimensions. This analysis not only facilitated a deeper understanding of wind turbine operational behaviors, but also allowed the identification of similar turbines across various temporal scales. In the final stage of the study, K-means clustering was utilized to identify outliers, which enabled the prediction of abnormal operational behaviors. The methodology proposed in this paper delivers a valuable clustering analysis technique for wind energy data, and provides significant insights for future data processing and anomaly prediction in wind turbine operations.